WO2008056649A1

WO2008056649A1 - Sound source position detector

Info

Publication number: WO2008056649A1
Application number: PCT/JP2007/071524
Authority: WO
Inventors: Shinichi Yoshizawa; Yoshihisa Nakatoh
Original assignee: Panasonic Corporation
Priority date: 2006-11-09
Filing date: 2007-11-06
Publication date: 2008-05-15
Also published as: DE602007011807D1; EP2090895A4; EP2090895B1; JP4177452B2; US20090285409A1; US8184827B2; JPWO2008056649A1; CN101512374B; CN101512374A; EP2090895A1

Abstract

A sound source position detector capable of detecting a sound source position of an extracted sound comprises at least two microphones; an analysis section (103) which performs a frequency analysis of mixed sound including a noise inputted to each of the microphones and outputs frequency signals; and an extraction section (105) which generates a plurality of sound source candidate positions of the extracted sound contained in the mixed sound, adjusts the time-axis of the frequency signals corresponding to the two microphones so as to eliminate a time difference between mixed sound reaching the two microphones from the sound source candidate position for each of the sound source candidate positions, determines frequency signals in which a difference distance indicating the difference degree of frequency signals between the two microphones out of the frequency signals corresponding to the two microphones after the adjustment of the time-axis is not greater than a threshold value, and extracts the sound source position of the extracted sound from the sound source candidate positions depending on the matching degree of the frequency signals corresponding to the two microphones out of the determined frequency signals.

Description

Specification

Sound source position detector

Technical field

[0001] The present invention relates to a sound source position detection device, a sound source position detection method, and a program for detecting a sound source direction or a sound source position of sound other than the noise by removing the influence of predetermined noise. In particular, the present invention relates to a vehicle-mounted sound source position detection device that detects the sound source direction or the sound source position of another vehicle existing around the host vehicle by removing the influence of the engine sound of the host vehicle. Background art

[0002] As a first conventional technique for detecting the position of a sound source, acoustic signals input to a plurality of microphones are mutually delayed and convolved to obtain a square error, and a delay when the square error value is minimized. There has been proposed one that specifies the sound source direction of the sound source signal based on the quantity! (For example, see Patent Document 1).

[0003] As a second conventional technique for detecting the sound source position of the extracted sound to be extracted from the mixed sound including noise, the fundamental speed component and the harmonics of the engine speed are detected by detecting the engine speed of the host vehicle. There has been proposed one that reduces the influence of the engine sound of the host vehicle by removing the wave component (see, for example, Patent Document 2).

[0004] FIG. 1 is a diagram for explaining a method of removing engine sound of a host vehicle according to a second conventional technique.

[0005] In the second prior art, first, a fundamental wave component fkHz of engine sound is determined using a sensor that detects the engine speed of the host vehicle. Next, 2 fkHz, 3 fkHz, and so on are calculated as harmonic components of the engine sound using the fundamental wave component f kHz. Figure 1 (a) shows the frequency signal of the waveform of the sound collected by the acoustic microphone! The horizontal axis shows the frequency and the vertical axis shows the signal level. Here, the signal in the frequency band of the fundamental wave component fkHz and harmonic components 2fkHz, 3fkHz, etc. (see Fig. 1 (b)) obtained earlier is removed from the frequency signal shown in Fig. 1 (a). As a result, as shown in Fig. 1 (c), a frequency signal is obtained by removing signals in the frequency band of the fundamental component fkHz and harmonic components 2fkHz, 3fkHz, ... of the engine sound. Using this frequency signal after removal, the sound source position of the vehicle around the vehicle Is required.

Patent Document 1: JP-A-6-83371 (Claim 2)

Patent Document 2: JP-A-5-85288 (Claim 2, Fig. 4)

Disclosure of the invention

Problems to be solved by the invention

[0006] However, in the first prior art, since there is no means for distinguishing the extracted sound to be extracted from the mixed sound including noise, it is difficult to detect the sound source position of the extracted sound from the mixed sound. there were.

[0007] In the second prior art, the fundamental frequency component and the harmonic component of the engine sound obtained from the engine speed are simply removed! /. Therefore, the effect of the engine sound of the vehicle cannot be completely removed. Specific examples will be described below.

FIG. 2 shows a spectrogram of the engine sound (idling sound) of the host vehicle input by the in-vehicle microphone. In the spectrogram, the vertical axis represents frequency and the horizontal axis represents time. Also, in this spectrogram, the light pressure level (signal level) is larger as the color is lighter and the sound pressure level is smaller as the color is darker.

[0009] According to the spectrogram shown in Fig. 2, not only the fundamental and harmonic components of the engine sound simply appear, but also the signal components appear complicatedly in other frequency bands. . One of the reasons for this is thought to be that the waveform was distorted by the influence of the hood when the engine sound was transmitted from the engine room to the microphone. For this reason, the influence of the engine sound of the vehicle cannot be completely removed by simply removing the fundamental and harmonic components of the engine sound. In order to make it easier to understand the spectrogram, Fig. 3 shows the binarized gray value of the spectrogram shown in Fig. 2.

[0010] The present invention has been made in view of such problems, and it is possible to detect a sound source direction or a sound source position of a sound other than the noise by removing the influence of predetermined noise. An object is to provide a position detection device and the like.

[0011] In particular, the influence of the engine sound of the host vehicle is removed, and the sound of other vehicles existing around the host vehicle is removed. The purpose is to provide a vehicle-mounted sound source position detection device that can detect the source direction or the sound source position.

Means for solving the problem

[0012] In order to achieve the above object, a sound source position detection device according to the present invention detects a sound source direction or a sound source position of an extracted sound to be extracted from a mixed sound including noise. An analysis means for analyzing a frequency of a mixed sound including noise input from each of a plurality of microphones arranged separately and outputting a frequency signal, and a plurality of sound sources of the extracted sound included in the mixed sound For each of the direction or sound source position candidates, (a) the time axis of the frequency signal corresponding to the plurality of microphones so that there is no time difference between the mixed sounds reaching the plurality of microphones from the sound source direction or sound source position candidates. (B) The difference distance indicating the degree of difference in the frequency signals among the plurality of microphones among the frequency signals corresponding to the plurality of microphones after the time axis adjustment. A frequency signal whose separation is less than or equal to a threshold value is obtained. (C) Among the obtained frequency signals, the frequency signals corresponding to the plurality of microphones are matched with each other from the sound source direction or sound source position candidates. Extraction means for extracting a sound source direction or a sound source position of the extracted sound, and when the extracted sound is input to the plurality of microphones, the threshold is the frequency signal obtained by frequency analysis of the extracted sound. The difference distance between a plurality of microphones is determined to be equal to or less than the threshold value, and the threshold value is determined when predetermined noise to be removed is input to the plurality of microphones. The difference distance between the plurality of microphones of the frequency signal obtained by frequency analysis of the predetermined noise is determined to be larger than the threshold value.

[0013] Thus, by comparing the difference distance by threshold processing, it is possible to distinguish between the noise to be removed and the extracted sound to be extracted, and therefore the frequency corresponding to the noise. By selectively removing the signal portion based on the frequency signal of the mixed sound input by the microphone, the influence of the noise can be removed and the sound source direction or the sound source position of the extracted sound can be detected.

[0014] FIG. 4A shows an example of a spectrogram in each microphone of noise (engine sound of the host vehicle). Figure 4B shows the frequency signal (powers in each microphone) of noise. An example of a spectrum is shown. Fig. 5A shows the spectrogram of each extracted sound (bike sound existing around the vehicle) at each microphone. Each of FIG. 5B and FIG. 5C shows an example of a frequency signal (power spectrum) in each microphone of the extracted sound. Each microphone is installed on the left and right sides of the front bumper of the vehicle.

[0015] FIG. 4A and FIG. 5A show spectrograms, with the vertical axis indicating frequency and the horizontal axis indicating time. In this spectrogram, the light pressure level (signal level) is higher as the color is lighter, and the sound pressure level is lower as the color is darker. Figures 4B, 5B, and 5C show the power spectrum, with the vertical axis representing the power spectrum and the horizontal axis representing the frequency. The two solid lines in the figure indicate the power spectrum at each microphone!

[0016] FIG. 4B shows a power spectrum at time A in FIG. 4A. Each microphone of the power spectrum at frequency A1 and frequency A2 is a frequency where noise (engine sound of the host vehicle) exists. It can be seen that the values at are significantly different. FIG. 5B shows the power spectrum at time B in FIG. 5A (the bike position is a point at a distance of 0 m from the host vehicle), and the frequency at which the extracted sound (the motorbike sound existing around the host vehicle) is present. It can be seen that the values at the microphones of the power spectrum at frequency B1 and frequency B2 are almost the same. Fig. 5C shows the power spectrum at time C in Fig. 5A (the bike position is a point 10m away from the vehicle), and there is an extracted sound (motorcycle sound that exists around the vehicle). It can be seen that the power spectrum values at frequencies C1 and C2 are almost the same for each microphone. From this, it can be seen that the noise and the extracted sound can be distinguished by comparing the difference distances by threshold processing. In this example, the noise is the engine sound of the host vehicle, and noise of different frequency signals is input to each microphone mounted on the host vehicle due to the influence of the bonnet or the difference in propagation distance. For this reason, the difference distance is increasing. On the other hand, the extracted sound is a motorcycle sound that exists at a position away from each microphone, and since the distance between the motorcycle position and each microphone can be approximated to be the same, the difference in sound pressure level due to distance attenuation is small, and further in the air A similar frequency signal of the motorcycle sound is input to each microphone that has low transmission distortion due to propagation. For this reason, the difference distance is reduced. [0017] Note that, when the spectrum power of the soot noise portion surrounded by the broken line in FIGS. 5B and 5C is observed, it can be seen that the values in the respective microphones are greatly different. From this, it can be seen that the power noise part can also be removed as noise.

[0018] Preferably, each of the microphones is divided by the distance between the noise source position and each of the microphones! /, The distance is large! /, And the object is divided by the object! When the ratio between the sound source position of the extracted sound and each of the microphones is larger than the ratio when the distance is smaller! ! /,

[0019] Thus, the degree of difference in the frequency signal (power spectrum) when noise reaches each microphone is determined by the frequency signal (no spectrum) when the extracted sound reaches each microphone. It can be made larger than the degree of difference (in the case of a distant sound source, it can be regarded as approximately zero). This means that the sound pressure level decays in inverse proportion to the distance (or the square of the distance), based on technical knowledge. When detecting distant extracted sound with microphones arranged at different distances near the noise (the ratio of the distance between the noise and each microphone is greater than 1), the distance between the distant extracted sound and each microphone is detected. This is true because the distance ratio can be approximated as 1.

[0020] Preferably, a substance having a transfer characteristic different between the microphones is used as a part of a medium through which the noise propagates to each microphone.

[0021] This makes it possible to increase the degree of difference in frequency signals (power spectrum and phase spectrum) when noise reaches each microphone, making it easy to set a threshold value. For example, when a substance with a different sound pressure level attenuation rate is sandwiched between noise and each microphone, the degree of difference in frequency signal (power spectrum) when noise reaches each microphone is increased. Can do. For example, it can be realized by sandwiching metals and plastics with different densities between a noise source and each microphone. In addition, when a substance with different phase characteristics is sandwiched between noise and each microphone, it is possible to increase the degree of difference in frequency signal (phase spectrum) when the noise reaches each microphone.

[0022] Preferably, the difference distance is a power spectrum of the frequency signal between the microphones. It is characterized by a different degree of tuttle.

[0023] This makes it possible to detect the sound pressure level in inverse proportion to the distance (or the square of the distance) when detecting far-field extracted sounds with microphones arranged at different distances in the vicinity of the noise. It is possible to directly determine the degree of difference in sound pressure level based on knowledge. In addition, when a substance with a different sound pressure level attenuation rate is sandwiched between noise and each microphone, the degree of difference in sound pressure level can be directly determined.

[0024] Preferably, the difference distance is obtained for each time interval having a predetermined time width.

[0025] Thus, the frequency signal force between the microphones can be removed by force S. Figure 6 shows the frequency signal (complex spectrum) of two microphones at 800 Hz. The dotted line shows the frequency signal in microphone A, and the solid line shows the frequency signal in microphone B. Here, since the frequency signal is a complex spectrum, the real part and the imaginary part are represented by separate graphs. In this example, the result when the sound source direction candidate is +30 degrees is shown. A portion 602 surrounded by a broken line in FIG. 6 is a frequency signal of a portion where the difference distance in the complex space for each time is small. From Fig. 6, it can be seen that when the difference distance is obtained for each time, the frequency signal power between the microphones, or the part that coincides only for a short time cannot be removed. However, when the difference distance is obtained for each time interval, it is possible to remove the portion that coincides only for a short time.

[0026] Preferably, the difference distance is a degree of difference between the frequency signals normalized by an average value of a power spectrum of the frequency signals of the plurality of microphones.

Accordingly, the difference distance does not depend on the sound pressure level, so that the threshold value can be easily set. For example, the threshold value can be easily set for the extracted sound having a high sound pressure level and the extracted sound having a low sound pressure level.

[0028] In order to achieve the above object, a vehicle-mounted sound source position detection device according to the present invention provides a sound source direction of a vehicle sound existing around the host vehicle from a mixed sound including noise generated by the host vehicle. Alternatively, a sound source position detection device that detects a sound source position and is arranged away from the host vehicle. Analyzing means for performing frequency analysis on a mixed sound including noise input from each of a plurality of microphones and outputting a frequency signal; and a plurality of vehicle sounds existing around the host vehicle included in the mixed sound For each of the sound source direction or sound source position candidates, (a) the frequency signal corresponding to the plurality of microphones is eliminated so that there is no time difference between the mixed sounds reaching the plurality of microphones from the sound source direction or sound source position candidate. (B) a frequency at which a difference distance indicating a degree of difference of the frequency signal among the plurality of microphones is equal to or less than a threshold value among frequency signals corresponding to the plurality of microphones after the time axis adjustment. (C) of the obtained frequency signals, the frequency signals corresponding to the plurality of microphones are matched according to the degree of coincidence between the plurality of sound source directions. Or extraction means for extracting the sound source direction or sound source position of the vehicle sound from the sound source position candidates, and the threshold value is determined when the vehicle sound is input to the plurality of microphones. The difference distance between the plurality of microphones of the frequency signal obtained by analyzing the frequency of the vehicle sound is determined to be equal to or less than the threshold value, and the threshold value is generated by the own vehicle in the plurality of microphones. When noise is input, the difference distance between the plurality of microphones of a frequency signal obtained by frequency analysis of the noise is determined to be larger than the threshold value.

As a result, the noise generated by the host vehicle and the vehicle sound existing around the host vehicle can be distinguished by comparing the magnitudes of the difference distances by threshold processing, and therefore, the frequency signal corresponding to the noise. By selectively removing this part based on the frequency signal of the mixed sound input by the microphone, the influence of the noise generated by the host vehicle (such as the engine sound of the host vehicle) is removed and it exists around the host vehicle. The ability to detect the sound source direction or position of the vehicle sound. For example, by using a material with a different density between the engine position and each microphone or a material with a different shape as the material of the bonnet, the difference distance of the frequency signal of the engine sound between each microphone is increased. Can do. In addition, when each microphone is embedded in a bumper, etc., a substance having different transfer characteristics between the microphones is used at the junction between the microphone and the bumper (the entire portion other than the portion where the microphone is outside the vehicle). In other words, the difference distance of the frequency signal of the engine sound between the microphones can be increased. And set the threshold easily That's the power S.

[0030] It should be noted that the present invention can be realized as a sound source position detecting device including such characteristic means, and a sound source position detecting method including the characteristic means included in the sound source position detecting device as a step. Or a program that causes a computer to execute the characteristic steps included in the sound source position detection method. And

Needless to say, such a program can be distributed via a recording medium such as a CD-ROM (Compact Disc-Read Only Memory) or a communication network such as the Internet.

The invention's effect

[0031] As described above, by comparing the difference distances, it is possible to distinguish between noise to be removed and extracted sound to be extracted. For this reason, the portion of the frequency signal corresponding to the noise is selectively removed based on the frequency signal of the mixed sound input by the microphone.

Therefore, it is possible to detect the sound source direction or the sound source position of the extracted sound by removing the influence of the noise. In particular, the effect of the engine noise generated by the subject vehicle, which is a problem when using a vehicle-mounted microphone, can be eliminated, so the effect is extremely large.

Brief Description of Drawings

[0032] FIG. 1 is a diagram for explaining a method for removing engine sound in the prior art.

FIG. 2 is a diagram showing an example spectrogram of the engine sound of the host vehicle.

[FIG. 3] FIG. 3 is a diagram in which the gray value of the spectrogram shown in FIG. 2 is binarized.

FIG. 4A is a diagram showing an example of a spectrogram of noise.

FIG. 4B is a diagram showing an example of a noise frequency signal (power spectrum).

FIG. 5A is a diagram showing an example of a spectrogram of extracted sound.

FIG. 5B is a diagram showing an example of a frequency signal (power spectrum) of the extracted sound.

FIG. 5C is a diagram showing an example of a frequency signal (power spectrum) of the extracted sound.

FIG. 6 is a diagram showing an example of a mixed sound frequency signal.

FIG. 7 is a block diagram showing an overall configuration of a sound source position detection device according to an embodiment of the present invention.

FIG. 8 is a diagram for explaining how to arrange microphones. FIG. 9 is a diagram showing an example of a configuration using substances having different transfer characteristics between the microphones of the host vehicle.

FIG. 10 is a diagram showing a configuration of a host vehicle including a radiator.

FIG. 11A is a diagram showing a spectrogram of the engine sound of the host vehicle.

FIG. 11B is a diagram showing a spectrogram of the engine sound of the host vehicle.

FIG. 12A is a diagram obtained by binarizing the grayscale value of the spectrogram shown in FIG. 11A.

FIG. 12B is a diagram obtained by binarizing the gray value of the spectrogram shown in FIG. 11B.

FIG. 13 is a diagram for explaining a sound source direction.

FIG. 14 is a flowchart showing an operation procedure of the sound source position detection device.

15A] FIG. 15A is a diagram showing an example of the result of analyzing the sound source direction of the engine sound of the host vehicle.

15B] FIG. 15B is a diagram showing an example of the result of analyzing the sound source direction of the engine sound of the own vehicle.

FIG. 16A is a diagram obtained by binarizing the gray value of the graph shown in FIG. 15A.

FIG. 16B is a diagram obtained by binarizing the gray value of the graph shown in FIG. 15B.

17A] FIG. 17A is a diagram showing an example of the result of analyzing the sound source direction of the engine sound of the own vehicle.

17B] FIG. 17B is a diagram showing an example of the result of analyzing the sound source direction of the engine sound of the own vehicle.

FIG. 18A is a diagram obtained by binarizing the gray value of the graph shown in FIG. 17A.

FIG. 18B is a diagram obtained by binarizing the gray value of the graph shown in FIG. 17B.

19A] FIG. 19A is a diagram showing an example of the result of analyzing the sound source direction of the engine sound of the host vehicle.

19B] FIG. 19B is a diagram showing an example of the result of analyzing the sound source direction of the engine sound of the host vehicle.

FIG. 20A is a diagram obtained by binarizing the gray value of the graph shown in FIG. 19A.

FIG. 20B is a diagram obtained by binarizing the gray value of the graph shown in FIG. 19B.

FIG. 21 is a diagram showing an example of a mixed sound frequency signal. FIG. 22 is a diagram showing an example of the frequency signal of the mixed sound.

23A] FIG. 23A is a diagram showing an example of the result of extracting the sound source direction using the mixed sound. FIG. 23B] FIG. 23B is a diagram showing an example of the result of extracting the sound source direction using the mixed sound.

FIG. 24A is a diagram obtained by binarizing the gray value of the graph shown in FIG. 23A.

FIG. 24B is a diagram obtained by binarizing the gray value of the graph shown in FIG. 23B.

25A] FIG. 25A is a diagram showing an example of the result of extracting the sound source direction using the mixed sound. FIG. 25B] FIG. 25B is a diagram showing an example of the result of extracting the sound source direction using the mixed sound.

FIG. 26A is a diagram obtained by binarizing the gray value of the graph shown in FIG. 25A.

FIG. 26B is a diagram obtained by binarizing the gray value of the graph shown in FIG. 25B.

Fig. 27A is a diagram showing an example of the result of extracting the sound source direction using the mixed sound. Fig. 27B is a diagram showing an example of the result of extracting the sound source direction using the mixed sound.

FIG. 28A is a diagram obtained by binarizing the gray value of the graph shown in FIG. 27A.

FIG. 28B is a diagram obtained by binarizing the gray value of the graph shown in FIG. 27B.

FIG. 29 is a diagram showing the sound source position of the engine sound of a motorcycle.

Explanation of symbols

100 Sound source position detector

101a Engine sound of own vehicle

101b Bike engine sound

102a ~ 102n microphone

103 Analysis Department

104 Shiki! BEST MODE FOR CARRYING OUT THE INVENTION

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

FIG. 7 is a block diagram showing a configuration of the sound source position detection apparatus in the embodiment of the present invention. Here, a case where the sound source position detection device according to the present invention is used for in-vehicle use will be described as an example.

The sound source position detection device 100 is a device that detects the sound source direction or the sound source position of other vehicles existing around the host vehicle, and includes two or more microphones 102a to 102n, an analysis unit 103, and an extraction unit 105. With. The microphone here refers to a device that converts an acoustic signal propagating through a medium (gas, solid, liquid) into an electrical signal.

[0037] Microphones 102a ~; 102η is a frequency analysis of the engine sound 101a (the sound to be removed) of the host vehicle whose time axis is adjusted so that the time difference to reach each of the microphones 102a ~ 102n is zero In the frequency signal, the difference distance between the microphone ports of the frequency signal whose time axis is adjusted is arranged to be larger than the threshold! /, Value 104 (threshold, value). The method for obtaining the “difference distance” will be described in detail later.

[0038] Fig. 8 is a diagram showing an example in which four microphones are arranged in the bumper in the front part of the host vehicle. In this example, the distance between the microphones A and B and the distance between the microphones C and D are 17.5 cm, and the distance between the microphones B and C is about lm.

[0039] The sound pressure level attenuates by 6 dB when the distance is doubled. Also, the distance between the vehicle's engine and the microphone is close. As a result, the ratio of the distance between the sound source and each microphone increases, and the sound pressure levels of the vehicle's engine sound input to the four microphones A, B, C, and D differ.

[0040] On the other hand, in the sound of a motorcycle engine that is far from the microphone, the ratio of the distance between the sound source and each microphone is approximately 1, and the sound pressure levels of the four microphones are approximately equal. I will be.

[0041] Further, it is considered that the engine sound of the host vehicle is distorted in the waveforms input to the four microphones due to the influence of the bonnet. In addition, intentionally using substances with different sound transmission characteristics at the connection between the four microphones and the host vehicle can be used to distort the waveforms input to the four microphones. [0042] In the engine sound transmission path from the engine position to each microphone, if there is a path that passes through the bonnet and a path that does not pass through the bonnet, or if the distance through the bonnet is different, the frequency signal received by each microphone Is different. Because of the difference in the density of substances present in the propagation path, such as the path passing through the bonnet does not pass through the bonnet where the density is high (for example, it propagates air), the attenuation rate of the engine sound This is because they are different.

[0043] Therefore, even if the distance of the engine sound transmission path from the engine position to each microphone is substantially equal, a substance having a different density or a substance having a different shape is required between the engine position and each microphone. By using the configuration, distortion can be caused in the waveform received by each microphone, and the difference distance can be increased.

[0044] In addition, when each microphone is embedded in a bumper or the like, transfer characteristics that differ between the microphones are present at the joint between the microphone and the bumper (the entire portion other than the portion where the microphone is outside the vehicle). Use a material that has the same material. Here, “different transmission characteristics” means that the acoustic impedance (density of the medium X velocity of the sound wave propagating in the medium) is different. Sound waves are reflected at their boundaries when propagating between materials with different acoustic impedances. The degree of reflection varies depending on the difference in acoustic impedance that increases as the difference in acoustic impedance between the propagating materials increases. Therefore, by using substances having different transfer characteristics between the engine position and each microphone, the waveform received by each microphone can be distorted and the difference distance can be increased.

[0045] Fig. 9 shows an example of a configuration using substances having different transfer characteristics between the microphones of the host vehicle. The host vehicle is equipped with an engine, metal that receives the engine sound generated by the engine force, and plastic between the engine and microphone B. It is assumed that engine sound (wave line) propagates from the engine to microphone A or microphone B! /. The engine sound propagates to the metal and to the plastic because the degree of reflection differs, so that distortion can occur between the waveforms received by microphone A and microphone B, increasing the difference distance. it can. Here we used metal and plastic, but different transmissions. If it is a substance which has a reachable characteristic, it has the effect that a difference distance can be enlarged. Here, the force described is an example with metal on one side and plastic on the other side. Even if only one side has metal or plastic, etc., one side has metal and plastic, and the other side has plastic, it has the same effect. .

FIG. 10 includes an engine, a microphone A, and a microphone B, as in FIG. In FIG. 10, a radiator is provided between the engine and the microphone A, and nothing is intentionally provided between the engine and the microphone B. A part of the engine sound transmitted from the engine to the microphone A is reflected due to the presence of the radiator. This is because the engine sound is reflected by the radiator due to the large difference in acoustic impedance between the liquid in the radiator and the air in which the engine sound has propagated. This reflection can cause distortion between the waveforms received by microphone A and microphone B, thereby increasing the difference distance. Therefore, there is a large difference between air and acoustic impedance between one microphone and the engine! /, And there is a substance between the other microphone and the engine. If the impedance changes, only the substance is sandwiched! /, The difference distance can be increased.

[0047] FIGS. 11A and 11B show the case where the difference distance between the microphone A and the microphone D in the engine sound 101a (idling sound) of the host vehicle is set to be larger than the threshold value 104. An example of the experimental results is shown. FIG. 11A shows a spectrogram of the engine sound 101a of the own vehicle input to the microphone A, and FIG. 11B shows a spectrogram of the engine sound 101a of the own vehicle input to the microphone D. The horizontal axis of each spectrogram indicates the time axis, and the vertical axis indicates the frequency axis. At each point of the spectrogram, the lighter the sound pressure level (signal level) the greater the color and the darker the color, the lower the sound pressure level.

[0048] When the spectrogram shown in FIG. 11A and the spectrogram shown in FIG. 11B are compared, the two spectrograms are totally different. In particular, in the region surrounded by the broken line, it can be seen with the naked eye that the frequency signal of the engine sound 101a of the own vehicle is distorted and different between the two spectrograms. In order to make it easy to grasp the spectrogram, the binarized values of the spectrogram shown in Fig. 11A and Fig. 11B are shown in Fig. 12A and Fig. 12B. Respectively.

[0049] The analysis unit 103 generates a mixed sound including the engine sound 101a of the own vehicle (the sound to be removed) and the engine sound 101b of the motorcycle (the sound to be extracted) input by the microphones 102a to 102n. Frequency analysis is performed to generate a frequency signal corresponding to each microphone.

[0050] The extraction unit 105 analyzes the sound source direction candidate force of the motorcycle engine sound 101b (the sound to be extracted), the microphones 102a to 102η, so that the time difference to reach each of 102η becomes zero. Adjusts the time axis of the analyzed frequency signal. In other words, the adjustment of the time axis means that the time of the frequency signal corresponding to each microphone 102a to 102n is eliminated so that there is no difference in arrival time when it is assumed that the motorcycle sound from the bike reaches each microphone 102a to 102n. Is to adjust the axis.

[0051] In addition, the extraction unit 105 removes the frequency signal in the portion where the differential distance force S between the microphones of the frequency signal with the time axis adjusted becomes larger than the S threshold 104, and uses the frequency signal in the remaining portion. Extract sound source direction.

[0052] Here, the sound source direction is indicated by a direction (angle) from 90 degrees to 90 degrees. FIG. 13 is a diagram for explaining the sound source direction. The sound source direction is 90 degrees for the right-hand direction and 90 degrees for the left-hand direction among the axes orthogonal to the straight direction when the straight and reverse directions of the host vehicle are 0 degrees.

[0053] Further, the threshold straight 104 is set such that the difference distance corresponding to the engine sound 101b of the motorcycle (the sound to be extracted) is equal to or less than the threshold value 104. From another viewpoint, the microphones 102a to 102η are arranged so that the difference distance corresponding to the engine sound 101b (the sound to be extracted) is equal to or less than the set threshold value 104. . In this example, the microphone arrangement, the difference distance corresponding to the engine sound 101a of the own vehicle in that arrangement, and the difference distance corresponding to the engine sound 101b of the bike are compared, and the sound source direction of the engine sound 101b of the bike is compared. The threshold 104 is experimentally determined so that can be detected.

As described above, by using the sound source position detection device 100, it becomes possible to remove the influence of the engine sound of the host vehicle and extract the direction of the sound source of the approaching motorcycle or the like to notify the driver. . For this reason, the sound source position detection device 100 is extremely useful as one function of safe driving support. Next, the operation of the sound source position detection device 100 configured as described above will be described.

FIG. 14 is a flowchart showing an operation procedure of the sound source position detection apparatus 100.

In this example, as shown in FIG. 8, four microphones A, B, C, and D are arranged in the bumper in the front part of the host vehicle. Further, for the microphone arrangement shown in FIG. 8, the threshold 104 is experimentally obtained from the difference distance corresponding to the engine sound 101a of the own vehicle and the difference distance corresponding to 10 lb of the engine sound of the motorcycle. (Step S901). In this example, the threshold 104 is set to 30%.

FIG. 15A and FIG. 15B show the results of analyzing the sound source direction of the engine sound 101a of the host vehicle by the conventional method using the microphone A and the microphone D shown in FIG. 8 without using threshold processing. An example is shown.

In the graphs shown in FIGS. 15A and 15B, the horizontal axis represents time, and the vertical axis represents the sound source direction (90 degrees to 90 degrees). In this graph, the darker the point, the higher the probability that a sound source exists in the direction of the sound source indicated by that point, and the lower the color, the lower the probability that a sound source exists in the sound source direction indicated by that point. Show me! /

[0060] In FIG. 15A, the frequency signal of both the portion where the difference distance corresponding to the engine sound 101a of the own vehicle becomes larger than the threshold 104 and the portion where it becomes smaller (the threshold value processing is performed). The result of analyzing the sound source direction by the difference in arrival time of the engine sound 101a of the own vehicle between the microphone A and the microphone D is shown. From FIG. 15A, it can be seen that the engine sound 101a of the host vehicle appears in a wide range of 90 to 90 degrees.

FIG. 15B shows the remaining frequency signal from which the difference distance corresponding to the engine sound 101a of the host vehicle is larger than the threshold value 104 (the threshold and value processing of the present invention). The result of analyzing the sound source direction by the difference in arrival time of the engine sound 101a of the vehicle between microphone A and microphone D is shown. Comparing Fig. 15A and Fig. 15B, it can be seen that the influence of the engine sound 101a of the own vehicle can be largely eliminated by removing the portion where the difference distance corresponding to the engine sound 101a of the own vehicle is larger than the threshold 104. . In order to make the graph easier to understand, the gray values in the graph shown in FIGS. 15A and 15B are binarized and shown in FIGS. 16A and 16B, respectively.

FIG. 17A and FIG. 17B use the microphone A and the microphone B shown in FIG. An example of the result of analyzing the sound source direction of the engine sound 101a is shown. The meanings of the graphs shown in FIGS. 17A and 17B are the same as those described in FIGS. 15A and 15B.

[0063] In FIG. 17A, the frequency signal of both the part where the difference distance corresponding to the engine sound 101a of the own vehicle becomes larger than the threshold value 104 and the part where it becomes smaller (the threshold value processing is performed). The result of analyzing the sound source direction by the difference in arrival time of the engine sound 101a of the own vehicle between the microphone A and the microphone B is shown. In FIG. 17B, the frequency signal of the remaining part from which the difference distance corresponding to the engine sound 101a of the own vehicle is larger than the threshold value 104 is removed (using the threshold / value processing of the present invention). The result of analyzing the sound source direction by the difference in arrival time of the engine sound 101a of the own vehicle between the microphone A and the microphone mouthphone B is shown. Comparing Fig. 17A and Fig. 17B, the effect of the engine sound 101a of the own vehicle can be largely eliminated by removing the portion where the difference distance corresponding to the engine sound 101a of the own vehicle is larger than the threshold 104. I understand. In order to make it easier to grasp the graph, binarized values of the graphs shown in FIGS. 17A and 17B are shown in FIGS. 18A and 18B, respectively.

FIGS. 19A and 19B show an example of the result of analyzing the sound source direction of the engine sound 101a of the host vehicle using the microphone C and the microphone D shown in FIG. The meanings of the graphs shown in FIGS. 19A and 19B are the same as those described in FIGS. 15A and 15B.

[0065] In FIG. 19A, the frequency signal of both the portion where the difference distance corresponding to the engine sound 101a of the own vehicle becomes larger than the threshold 104 and the portion where it becomes smaller is used (the threshold value processing is performed). The result of analyzing the sound source direction based on the arrival time difference of the engine sound 101a of the own vehicle between the microphone C and the microphone D is shown. In FIG. 19B, a microphone is used by using the frequency signal of the remaining part from which the difference distance corresponding to the engine sound 101a of the own vehicle is larger than the threshold value 104 (using threshold and value processing). The result of analyzing the sound source direction by the difference in arrival time of the engine sound 101a of the own vehicle between C and microphone D is shown. Comparing FIG. 19A and FIG. 19B, it is possible to largely eliminate the influence of the engine sound 101a of the own vehicle by removing the portion where the difference distance corresponding to the engine sound 101a of the own vehicle is larger than the threshold 104. Recognize. In order to make the graph easier to understand, the gray values in the graphs shown in FIGS. 19A and 19B are binarized and shown in FIGS. 20A and 20B. Each is shown in 20B.

From this, it can be seen that the microphone arrangement shown in FIG. 8 is an arrangement in which the difference distance corresponding to the engine sound 101a of the host vehicle is larger than the threshold value 104. However, as can be seen from FIG. 15B, FIG. 17B, and FIG. 19B, in this example, the influence of the engine sound 101a of the own vehicle has not been completely removed. In this example, the predetermined noise to be removed is used. And here, the direction of the sound of the engine sound of the bike is not removed! /, Like! /, Set the value! /,

[0067] It should be noted that one microphone is placed closer to the engine so that the ratio of the distance between the engine of the vehicle and each microphone (the ratio when the distance is large divided by the distance is small) is large. By placing one microphone far away from the engine at the position, it is possible to increase the difference distance from the engine sound of the vehicle and facilitate the threshold setting. Also, by using a material with a different density or a material with a different shape between the engine position and each microphone as the bonnet material, the difference distance to the engine sound of the vehicle can be increased, and a threshold value can be set. Can facilitate

[0068] Next, the microphones 102a to 102n input mixed sounds including the engine sound 101a of the own vehicle (the sound to be removed) and the engine sound 101b of the motorcycle (the sound to be extracted) (sound). Tape S 902).

[0069] Next, the analysis unit 103 performs frequency analysis on the mixed sound input by the microphones 102a to 102n and creates a frequency signal corresponding to each microphone (step S903). Here, a mixed sound with a sampling frequency of 44.1 kHz is created with a discrete Fourier transform window length of 128 points in a frequency band of 0 to 2000 Hz at 20 Hz intervals! .

[0070] Next, the extraction unit 105 performs analysis so that the time difference from the candidate of the sound source direction of the engine sound 101b (the sound to be extracted) of the motorcycle to each of the microphones 102a to 102η becomes zero. 103 adjusts the time axis of the frequency signal analyzed. Further, the extraction unit 105 extracts the direction of the sound source by using the remaining frequency signal from which the frequency signal in which the difference distance between the microphones of the frequency signal adjusted in time axis is larger than the threshold 104 is removed. (Step S904).

[0071] In FIG. 21, the time difference between the frequency signal in the 800 Hz frequency band from the sound source direction candidate (+50 degrees) to each of the microphones for the mixed sound input by microphone A and microphone B is shown in FIG. A frequency signal with the time axis adjusted to zero is shown.

A dotted line indicates a frequency signal in microphone A, and a solid line indicates a frequency signal in microphone B. The horizontal axis indicates the time axis, and the vertical axis indicates the amplitude. Here, since the frequency signal is a complex signal, the real part and the imaginary part are shown in separate graphs.

Here, an example of how to obtain the difference distance will be described. In this example, in the time interval shown in FIG. 21 (time interval having a predetermined time width), the proportion of the time portion 300 in which the difference value of the frequency signal at each time is greater than the predetermined value occupies the entire time interval. Is the “difference distance”. The predetermined value here is obtained experimentally. Also, the difference value here is a value normalized by the magnitude of the amplitude. For example, regarding the degree of distortion of the amplitude value, it is considered that the larger the amplitude value, the larger the absolute amount of the distorted amplitude value. For example, if the sound pressure level is attenuated by half, and the amplitude value before attenuation is 10, the amplitude value after attenuation is 5 and the difference is 5. When the amplitude value before attenuation is 2, the amplitude value after attenuation is 1 and the difference is 1. Therefore, the larger the amplitude value, the larger the absolute amount of the distorted amplitude value. In this way, it is considered that the influence of the magnitude of the amplitude can be eliminated by using the difference value normalized by the magnitude of the frequency signal. In this example, the “difference value” is obtained by the following equation. As a result, since the difference distance does not depend on the sound pressure level, for example, it is easy to set the threshold and the value for the large sound pressure level V, the small amount of the extracted sound and the sound pressure level, and the extracted sound. Can be done.

[0074] country

[0075] The difference value is the average of the power spectrum of the frequency signal Since it is normalized by the value, it is possible to set the threshold straight without depending on the sound pressure level depending on the frequency to be analyzed. Where X and y are the frequencies in microphone A

A A

Represents the real and imaginary parts of the signal at times, where x and y are microphones B

B B

Represents the values of the real part and the imaginary part of the corresponding time of the frequency signal in FIG. In addition, the predetermined value in this example is 0.5, and the threshold value 104 is 30%. Since the difference value is normalized by the size of the spectrum, it is not necessary to consider the magnitude of the power spectrum with respect to the frequency band. For this reason, the predetermined value is the same for all frequency bands. A portion 300 surrounded by a broken line in FIG. 21 is a frequency signal of a portion where the difference value is larger than a predetermined value. From Fig. 21, it can be seen that when the time axis of the frequency signal is adjusted assuming that the sound source position exists in the direction of +50 degrees, the portion 300 surrounded by the broken line occupies a large proportion of the entire time interval. If the ratio, which is the difference distance, is larger than the threshold 104, which is 30%, the direction of the sound source is determined by removing the frequency signal for the entire time interval. At this time, it is interpreted that there is no sound source to be extracted in the direction of +50 degrees with respect to the 800 Hz frequency band.

[0076] In Fig. 22, the frequency signal in the frequency band of 800Hz is reached for each of the microphones from other sound source direction candidates (130 degrees) for the mixed sound input by microphone A and microphone B. The frequency signal is shown with the time axis adjusted so that the time difference is zero.

A dotted line indicates a frequency signal in the microphone A, and a solid line indicates a frequency signal in the microphone B. The horizontal axis indicates the time axis, and the vertical axis indicates the amplitude. Since the frequency signal is a complex signal here, the real part and the imaginary part are represented by separate graphs.

Again, in the same time interval (time interval having a predetermined time width) as the time interval shown in FIG. 21, the difference distance is obtained using the same definition. A portion 302 surrounded by a broken line is a frequency signal of a portion whose difference value is larger than a predetermined value. As can be seen from FIG. 22, when the time axis of the frequency signal is adjusted assuming that the sound source position exists in the direction of −30 degrees, the portion 302 surrounded by the broken line occupies a small proportion of the entire time interval. If the percentage of this difference distance is less than 30%, which is the threshold 104, the frequency signal for this entire time interval The direction of the sound source is obtained without removing the signal. In this example, when obtaining the sound source direction, all the correlation values (degrees of coincidence) between the microphones in the portion where the difference value at each time in this time interval is less than or equal to a predetermined value are not removed. If the cross-correlation value in the sound source direction (angle) has the largest value, it is determined that there is a sound source to be extracted in the direction of 30 degrees with respect to the 800 Hz frequency band. For example, assuming that candidates for sound source direction remain at 30 degrees, 40 degrees, and 50 degrees for the frequency band of 800 Hz, the sound source direction corresponding to the maximum of the cross-correlation values in each candidate Are identified as the sound source direction of the extracted sound. Note that the sound source direction candidate corresponding to the correlation value exceeding the second predetermined threshold may be specified as the sound source direction of the extracted sound. In this case, it can cope with the direction of multiple sound sources. Further, the sound source direction candidate corresponding to the one having the maximum correlation value with respect to the axis of the sound source direction may be specified as the sound source direction of the extracted sound. Also, as another degree of coincidence, all directions where the difference distance is equal to or less than the threshold may be set as the sound source direction. The direction in which the difference distance takes the minimum value or the minimum value may be set as the sound source direction.

[0079] FIG. 23A, FIG. 23B, FIG. 25A, FIG. 25B, FIG. 27A and FIG. 27B show 100 frequency bands in which the direction of the sound source obtained by the conventional method and the method according to the present invention is included in 0 to 2000 Hz. A histogram is shown for. Figures 23A and 23B show the results using microphone A and microphone mouthphone D, Figures 25A and 25B show results using microphone A and microphone B, and Figures 27A and 27B show microphone C and microphone D. The results using are shown.

[0080] The horizontal axis represents the time axis, and the vertical axis represents the sound source direction. The darkness of the color represents the frequency, and the darker the part, the higher the frequency. Figures 23A, 25A, and 27A show the results of the prior art in which the sound source direction was extracted by using both threshold, larger than value, part and small, and part (without using threshold and value processing). It is. Figures 23B, 25B, and 27B show the results of the present invention in which the direction of the sound source was extracted using the threshold V, the frequency signal larger than the value, and the frequency signal from which the portion was removed (using the threshold and value processing). It is. To make it easier to grasp the graph, the gray values in the graphs shown in FIGS. 23A, 23B, 25A, 25B, 27A, and 27B are binarized, as shown in FIGS. 24A, 24B, and 24 It is shown in FIG. 26A, FIG. 26B, FIG. 28A and FIG. 28B, respectively. FIG. 29 schematically shows the sound source position of the motorcycle engine sound 10 lb corresponding to the time. The direction of the sound source seen from microphone A and microphone B is about 0 degrees per 3 seconds (the direction of the sound source seen from microphone C and microphone D is about +30 degrees), and microphone A and microphone are taken every 5 seconds. The direction of the sound source seen from B is about +90 degrees, and then goes to 0 degrees.

FIG. 15A, FIG. 17A, and FIG. 19A show the sound source direction of the engine sound 101a (idling sound) of the host vehicle. The results of FIGS. 23B, 25B, and 27B show the results of the present invention. It can be seen that the sound source direction of the engine sound 101a of the own vehicle is removed by processing the threshold value as a point, and the sound source direction of the engine sound 101b of the motorcycle can be extracted. 19A and 27B, the influence of the engine sound 101a of the own vehicle can be removed even though the sound source direction of the engine sound 101a of the own vehicle and the sound source direction of the engine sound 101b of the motorcycle overlap. You can see that

As described above, it can be seen that the noise to be removed and the extracted sound to be extracted can be distinguished by performing threshold processing by comparing the difference distances. For this reason, the portion of the frequency signal corresponding to noise is selectively removed based on the frequency signal of the mixed sound input by the microphone, thereby removing the influence of noise and accurately detecting the sound source direction of the extracted sound. That power S. In particular, the effect of the engine sound generated by the host vehicle, which is a problem when using a vehicle-mounted microphone, can be eliminated, so the effect is extremely great.

[0084] In this example, the difference distance is occupied over the entire time interval of the time portion where the difference value of the amplitude value of the frequency signal at each time is greater than the predetermined value in the time interval having a predetermined time width. However, as another difference distance, a temporally continuous amplitude shape is represented as a vector in a time interval with a predetermined time width, and the difference value between the outer loops is defined as the difference distance. You can also. Also in this case, the difference distance is obtained in a time interval having a predetermined time width. For this reason, it is possible to remove the coincident part for a short time due to the amplitude fluctuation of the frequency signal between the microphones. In this example, a force power spectrum using a difference distance in a complex spectrum space or a difference distance of a phase spectrum can also be used.

As described above, by calculating the difference distance for each time interval having a time width, It is possible to remove the portion where the amplitude value of the frequency signal between the signals coincides only for a short time due to the amplitude fluctuation of the frequency signal. In addition, the influence of distortion of the amplitude value due to the mixing of multiple sounds can be eliminated.

[0086] Also, since the motorcycle moves, the sound pressure level increases when approaching the host vehicle with a low sound pressure level in the distance from the host vehicle. Therefore, the average of the spectrum of frequency signals between the microphones is averaged. It is considered effective to use the difference distance normalized by the value.

The sound source position can be obtained by triangulation from the extraction result from D. It is also possible to directly find the sound source position by adjusting the arrival time difference from the sound source position candidate to each microphone and comparing the degree of coincidence.

[0088] As described above, since the noise to be removed and the extracted sound to be extracted can be distinguished by comparing the difference distances and thresholding them, the frequency corresponding to the noise can be distinguished. By selectively removing the several signal parts based on the frequency signal of the mixed sound input by the microphone, it is possible to remove the influence of noise and detect the sound source direction of the extracted sound. In particular, the effect of the engine sound generated by the host vehicle, which is a problem when using an on-vehicle microphone, can be eliminated, and the effect is extremely large.

It should be noted that the threshold value 104 is determined based on the result of the extraction unit 105 (for example, the degree of removal of the sound source direction of the engine sound 101a of the own vehicle or the result of extraction of the sound source direction of the noisy engine sound 101b). The value of may be further adjusted.

[0090] It should be noted that since the noise signal and wind noise are considered to have different frequency signals in each microphone, it is considered that these sounds can also be removed as noise. Also, even when different sounds are input between the microphones, the frequency signals are different for each microphone, so the sound in this case can also be removed as noise, preventing the wrong sound source direction or sound source position from being output. .

Industrial applicability

The sound source position detection device according to the present invention is a safe driving support device that detects an approaching vehicle by sound (for example, by removing the influence of the engine sound of the host vehicle), (for example, a shadow of a motor sound of a camera). A security camera that automatically directs the camera to the sound source position (with the sound removed), and can collect sound with directivity in the direction of the speaker (removing the influence of the other party's utterance in the video conference) It can be applied to a wide range of products such as video conferencing equipment, and its practical value is extremely high.

Claims

The scope of the claims

[1] A sound source position detection device for detecting a sound source direction or a sound source position of an extracted sound to be extracted from a mixed sound including noise,

Analyzing means for frequency-analyzing a mixed sound including noise inputted from each of a plurality of microphones arranged apart from each other and outputting a frequency signal;

For each of a plurality of sound source directions or sound source position candidates of the extracted sound included in the mixed sound, (a) the time difference of the mixed sound reaching the plurality of microphones from the sound source direction or sound source position candidate is eliminated. (B) adjusting the time axis of the frequency signal corresponding to the plurality of microphones, and (b) the degree of difference of the frequency signal between the plurality of microphones among the frequency signals corresponding to the plurality of microphones after the time axis adjustment. (C) Among the obtained frequency signals, a candidate of a sound source direction or a sound source position is determined based on the degree of coincidence of the frequency signals corresponding to the plurality of microphones. Extraction means for extracting the sound source direction or the sound source position of the extracted sound from

When the extracted sound is input to the plurality of microphones, the threshold is such that the difference distance between the plurality of microphones of the frequency signal obtained by frequency analysis of the extracted sound is equal to or less than the threshold. Is defined as

In addition, when the predetermined noise to be removed is input to the plurality of microphones, the threshold value is the frequency signal obtained by frequency analysis of the predetermined noise between the plurality of microphones. The sound source position detection device, wherein the difference distance is determined to be larger than the threshold value.

[2] Each microphone has a distance between the noise source position and each microphone! /, The distance is large! /, And the distance is small! / The distance between the sound source position of the extracted sound and each of the microphones! /, The distance is large! /, And the distance is larger than the ratio when the distance is divided by the smaller distance. The sound source position detection device according to claim 1, wherein

[3] A substance having different transmission characteristics between microphones is used as a part of a medium through which the noise propagates to each microphone. The sound source position detection device according to claim 1, wherein:

[4] The difference distance is the degree of difference in the power spectrum of the frequency signal between the microphones.

The sound source position detection device according to claim 1, wherein:

[5] The difference distance is obtained for each time interval having a predetermined time width.

The sound source position detection device according to claim 1, wherein:

[6] The difference distance is a degree of difference between the frequency signals normalized by an average value of a power spectrum of the frequency signals of the plurality of microphones.

The sound source position detection device according to claim 1, wherein:

[7] A sound source position detection device that detects a sound source direction or a sound source position of a vehicle sound existing around the own vehicle from a mixed sound including noise generated by the own vehicle, and is disposed apart from the own vehicle. Analyzing means for frequency-splitting the mixed sound including noise input from each of the plurality of microphones and outputting a frequency signal;

For each of a plurality of sound source directions or sound source position candidates for vehicle sound existing in the vicinity of the host vehicle included in the mixed sound, the mixed sound that reaches the plurality of microphones from the sound source direction or sound source position candidates. The time axis of the frequency signal corresponding to the plurality of microphones is adjusted so as to eliminate the time difference of (b), and (b) the frequency signal corresponding to the plurality of microphones after the time axis adjustment is set between the plurality of microphones. A frequency signal having a difference distance indicating a degree of difference between frequency signals equal to or less than a threshold value is obtained. Extraction means for extracting the sound source direction or sound source position of the vehicle sound from the sound source direction or sound source position candidates;

When the vehicle sound is input to the plurality of microphones, the threshold value is such that the difference distance between the plurality of microphones of a frequency signal obtained by frequency analysis of the vehicle sound is equal to or less than the threshold value. Is defined as

In addition, when the noise generated by the host vehicle is input to the plurality of microphones, the threshold value is the threshold value of the difference distance between the plurality of microphones of a frequency signal obtained by frequency analysis of the noise. , Determined to be larger than the value! / A vehicle-mounted sound source position detection device characterized by the above.

[8] A sound source position detection method for detecting a sound source direction or a sound source position of an extracted sound to be extracted from a mixed sound including noise,

An analysis step of frequency-analyzing a mixed sound including noise input from each of a plurality of spaced microphones and outputting a frequency signal;

For each of a plurality of sound source directions or sound source position candidates of the extracted sound included in the mixed sound, (a) the time difference of the mixed sound reaching the plurality of microphones from the sound source direction or sound source position candidate is eliminated. (B) adjusting the time axis of the frequency signal corresponding to the plurality of microphones, and (b) the degree of difference of the frequency signal between the plurality of microphones among the frequency signals corresponding to the plurality of microphones after the time axis adjustment. (C) Among the obtained frequency signals, a candidate of a sound source direction or a sound source position is determined based on the degree of coincidence of the frequency signals corresponding to the plurality of microphones. Extracting a sound source direction or a sound source position of the extracted sound from

In addition, when the predetermined noise to be removed is input to the plurality of microphones, the threshold value is the frequency signal obtained by frequency analysis of the predetermined noise between the plurality of microphones. The sound source position detection method, wherein the difference distance is determined to be larger than the threshold value.

[9] A program for detecting a sound source direction or a sound source position of an extracted sound to be extracted from a mixed sound including noise,

For each of a plurality of sound source directions or sound source position candidates of the extracted sound included in the mixed sound, (a) the time difference of the mixed sound reaching the plurality of microphones from the sound source direction or sound source position candidate is eliminated. Frequency corresponding to the plurality of microphones After adjusting the time axis of the numerical signal, and after adjusting the axis adjustment of the time ((bb)) The frequency frequency signal signal of the frequency frequency signal between the plurality of microphones in front of the frequency signal signal corresponding to The difference frequency distance indicating the degree of difference in the difference of the distance difference distance is determined to obtain a frequency signal that is less than or equal to the threshold value, and (( cc)) The frequency signal of the frequency signal corresponding to the plural number of microphone signals before the above-mentioned frequency signal of the determined frequency signal. Depending on the degree of match of each other, depending on the degree of match of the sound source, it remains in the direction of the sound source source or from the middle candidate of the sound source source position candidate Extraction extraction Let the computer execute the extraction step to extract the sound source source position or the sound source source position of the sound. ,,

The above-mentioned threshold value is obtained when the above-mentioned extracted extracted sound is input / inputted to a plurality of microphones. In addition, before the frequency frequency signal obtained by analyzing the extracted and extracted sound by frequency frequency analysis, between the plurality of microphone chlorophones. Before the difference difference distance is determined to be less than or equal to the above threshold value,

However, the above-mentioned threshold value is a predetermined value that is to be removed and removed from the plurality of mama chlorophophones. In the case where the noise signal of the noise is input / input force, the frequency signal of the frequency signal obtained by analyzing the noise signal of the predetermined frequency is analyzed for the frequency frequency. In front of the number, the difference between the plurality of mama's mouths is larger than the threshold value. It has been determined to become a *